An Improved Local Multi-scale Retinex Algorithm Based on Illuminance Image Segmentation
Xintong Zhang · Dianzi xuebao · 2010
The linear weighted average method is adopted to integrate the results of different scale Retinex Algorithms,but it did not stress the merit of each scale in maintaining the details and the color fidelity effectively.So a novel Multi-Scale Retinex Algorithm is proposed in this paper.It segments original image into four illuminance regions by involving the parameterized LIP model.Then each region is enhanced by Retinex Algorithm whose scale is selected according to the illuminance difference.At last,based on proportionality factor,the enhanced subimages are fused to enhance the intact image.Experimental results show that the new algorithm is better than the traditional MSR method in both details and color fidelity.